What it does

GitHub Copilot is GitHub's AI coding assistant. It does code completion, chat, agentic multi-file edits, autonomous background work and code review across the editor, github.com, the terminal and mobile.

Key facts

• Runs across several surfaces: IDEs (VS Code, Visual Studio, JetBrains, Neovim, Eclipse, Xcode), github.com (chat, agents panel, code review), the Copilot CLI, and GitHub Mobile.
• Copilot CLI went GA on 25 February 2026 (public preview from September 2025) as a terminal-native agent that plans, builds and reviews without leaving the shell.
• Model choice is multi-vendor: current families span Anthropic Claude (Opus/Sonnet), OpenAI GPT and Codex, Google Gemini and xAI Grok. Exact model versions move monthly.
• The autonomous coding agent reached GA on 25 September 2025; Copilot code review reached GA in April 2025.
• A remote-control feature (GA 18 May 2026) lets you steer a CLI session from the web, GitHub Mobile, VS Code or JetBrains via QR code.
• Copilot SDK went GA on 2 June 2026 for building custom agents on the same stack.
• It is a distinct product from Microsoft 365 Copilot and Copilot Studio — GitHub Copilot is developer-facing and billed through GitHub, not the M365 estate.

When to use / skip

Reach for GitHub Copilot whenever the work touches source code, repositories, issues or CI. It is the developer arm of the Microsoft Copilot family, not a general office assistant. If the user wants document drafting, Teams summaries or Power Platform work, that is M365 Copilot or Copilot Studio territory — don't conflate them just because the branding overlaps.

Configuration decisions

• Decide which surfaces you sanction — IDE only, or IDE plus github.com agents and CLI. Each opens different data-flow and governance questions.
• Pick a default model policy, or let auto-selection run. Locking models gives predictable cost and behaviour; leaving it open gives developers newer models sooner.
• Choose whether the coding agent and CLI are in scope; both need explicit enablement on Business/Enterprise.
• Work out identity: GitHub org membership and SSO drive seat assignment, so sort the GitHub Enterprise Cloud setup first.

Gotchas

• "Copilot" means at least four different Microsoft products. Be precise in scoping calls — GitHub Copilot is the one for engineers.
• Feature availability differs by surface and plan. A capability in VS Code may not exist yet on github.com or in the CLI.
• Model names and defaults change fast. Anything you write down about "the current model" is stale within weeks.
• JetBrains, Eclipse and Xcode surfaces usually trail VS Code for new features — set expectations if the team isn't on VS Code.

Consultant notes

• Position it as the engineering-team purchase, funded and governed separately from the M365 Copilot rollout. Different buyer, different licence, different admin plane.
• The multi-model story is a genuine selling point against single-vendor assistants — teams can switch between Claude and GPT in the same editor.
• Adoption is broad (roughly 90% of the Fortune 100 use the coding agent), so the "is it enterprise-ready" objection is largely spent.
• Scope surfaces deliberately. Turning everything on day one gives you a governance headache; start with completions and chat, add agents once policy is settled.

---

Check the model list before every client demo — it will have changed since you last looked.